[人工智能]ML-GCN遇到的问题（未解决）

Traceback (most recent call last):
  File "demo_coco_gcn.py", line 76, in <module>
    main_coco()
  File "demo_coco_gcn.py", line 73, in main_coco
    engine.learning(model, criterion, train_dataset, val_dataset, optimizer)
  File "/userhome/gyt/ML-GCN/engine.py", line 202, in learning
    self.train(train_loader, model, criterion, optimizer, epoch)
  File "/userhome/gyt/ML-GCN/engine.py", line 253, in train
    self.on_end_epoch(True, model, criterion, data_loader, optimizer)
  File "/userhome/gyt/ML-GCN/engine.py", line 335, in on_end_epoch
    map = 100 * self.state['ap_meter'].value().mean()
  File "/userhome/gyt/ML-GCN/util.py", line 225, in value
    ap[k] = AveragePrecisionMeter.average_precision(scores, targets, self.difficult_examples)
  File "/userhome/gyt/ML-GCN/util.py", line 251, in average_precision
    precision_at_i /= pos_count
ZeroDivisionError: float division by zero

当把数据的标签改为119×119之后，train.sh遇到上述问题，原因可能是prectision的准确率太低，导致计算AP的时候pos_count=0.